Zhuohang Bian

Zhuohang Bian

Peking University

I’m Zhuohang Bian (εžε“θˆͺ), a Ph.D. student at the School of Integrated Circuits, Peking University. Before that, I studied Computer Science and Technology at Beihang University.

My research interests center on efficient LLM inference systems, multi-agent workloads, computer architecture, and hardware-software co-design. I am currently working with Prof. Youwei Zhuo on agent systems. If you are interested in potential collaborations or would like to discuss anything, please feel free to reach out by email. Here is my CV.

πŸ”₯ News

  • 2026.08: πŸŽ‰ TokenCake was accepted to EuroSys 2027.
  • 2026.08: 🎀 I am serving as the chair of the APPT tutorial T2: Graph.hls: A Compiler Framework for Composable Graph Accelerator Design.
  • 2026.07: πŸŽ‰ TokenSimV2: Accurate and Fast Modeling for LLM Inference on GPUs was accepted to APPT.
  • 2026.06: πŸ† I was awarded the title of Outstanding Graduate of Beijing (εŒ—δΊ¬εΈ‚δΌ˜η§€ζ―•δΈšη”Ÿ).
  • 2026.04: πŸš€ We released TokenDance, a system for collective KV cache sharing in multi-agent LLM serving.
  • 2025.10: πŸ’» I completed a systems internship at SenseTime, with six pull requests merged into LightLLM.
  • 2025.09: πŸŽ“ I was admitted to the Ph.D. program at the School of Integrated Circuits, Peking University.

πŸ“ Publications

  • TokenCake: A KV-Cache-centric Serving Framework for LLM-based Multi-Agent Applications
    Zhuohang Bian, Feiyang Wu, Zhuoran Li, Teng Ma, Youwei Zhuo
    EuroSys 2027
    [paper]

  • TokenDance: Scaling Multi-Agent LLM Serving via Collective KV Cache Sharing
    Zhuohang Bian, Feiyang Wu, Chengrui Zhang, Hangcheng Dong, Yun Liang, Youwei Zhuo
    Preprint
    [paper]

  • A Full-Stack Characterization of High-Bandwidth Flash for KV-Centric LLM Serving
    Zhuoran Li, Zhuohang Bian, Xin Huang, Yibo Zhao, Guangyu Sun, Youwei Zhuo
    Preprint
    [paper]

  • TokenStack: A Heterogeneous HBM-PIM Architecture and Runtime for Efficient LLM Inference
    Zhuoran Li, Zhuohang Bian, Zihao Huang, Yibo Zhao, Xueqi Li, Guangyu Sun, Youwei Zhuo
    Preprint
    [paper]

  • Graph.hls: A Compiler Framework for Composable Graph Accelerator Design
    Feiyang Wu, Xuxiao Yang, Zhuohang Bian, Jing Wang, Ruifan Xu, Guangyu Sun, Yun Liang, Youwei Zhuo
    ISCA 2026
    [paper]

  • TokenSim: Enabling Hardware and Software Exploration for Large Language Model Inference Systems
    Feiyang Wu, Zhuohang Bian, Guoyang Duan, Tianle Xu, Junchi Wu, Teng Ma, Yongqiang Yao, Ruihao Gong, Youwei Zhuo
    APPT 2025
    [paper]

  • MammoExpert: Benchmarking Chain-of-Thought Reasoning in Mammography Diagnosis
    Di Dai, Bo Liu, Youcheng Li, Haojun Yu, Zhuohang Bian, Quanlin Wu, Dong Wang, Sichen Meng, Hongye Xuan, Zijie Lan, Shenda Hong, Liwei Wang
    KDD 2026
    [paper]

  • Empirical Study on Robustness and Resilience in Cooperative Multi-Agent Reinforcement Learning
    Simin Li, Zihao Mao, Hanxiao Li, Zonglei Jing, Zhuohang Bian, Jun Guo, Li Wang, Zhuoran Han, Ruixiao Xu, Xin Yu, Chengdong Ma, Yuqing Ma, Bo An, Yaodong Yang, Weifeng Lv, Xianglong Liu
    NeurIPS 2025
    [paper] [code]

πŸ”¬ Research Experience

  • Peking University, School of Integrated Circuits
    Research Assistant, Prof. Youwei Zhuo’s Group, Feb 2025 - Present
    I lead research on KV-cache-centric serving for multi-agent LLM applications. TokenCake combines agent-aware scheduling and memory management, while TokenDance introduces collective KV cache sharing and block-sparse diff storage.

  • State Key Laboratory of Complex & Critical Software Environment, Beihang University
    Research Assistant, Prof. Xianglong Liu’s Group, Jul 2024 - Jan 2025
    I built AMB, a robustness benchmark covering four real-world environments, 13 uncertainty types, 15 key hyperparameters, and 82,620 experiments for cooperative multi-agent reinforcement learning.

  • Sino-German Joint Software Institute, Beihang University
    Research Assistant, Prof. Zhongzhi Luan’s Group, Sep 2023 - May 2024
    I optimized multi-head attention for the MT-3000 CPU-DSP heterogeneous processor with matrix tiling, DMA double buffering, communication optimization, and head-level operator scheduling.

πŸ’Ό Industry Experience

  • SenseTime - LightLLM
    Systems Research Intern, Jun 2025 - Oct 2025
    I worked on prefill-decode disaggregated scheduling, memory-aware routing, cross-GPU KV cache migration, and MinerU2 multimodal model support. Six public pull requests were merged: #944, #970, #999, #1011, #1024, and #1034.

πŸ› οΈ Selected Projects

  • Out-of-order LoongArch Processor
    I implemented key backend modules for a 13-stage dual-issue superscalar processor supporting LoongArch32R. The processor boots Linux, reached 80 MHz on FPGA, and placed fourth nationally in the 2024 Loongson Cup CPU Track. [code]

  • Compiler Course Infrastructure
    I developed a compiler course project that was adopted as a teaching example in national compiler-course teacher training. [code]

πŸ… Honors and Awards

  • Outstanding Graduate of Beijing β€” Beijing Municipal Education Commission, 2026
  • First Prize, China Undergraduate Mathematical Contest in Modeling β€” National Level, China Society for Industrial and Applied Mathematics, Nov 2024 (Team Award)
  • 4th Place / Second Prize, Loongson Cup CPU Track β€” National College Student Computer System Ability Competition, Aug 2024 (Team Award)
  • First Prize (Meritorious Winner), International Mathematical Contest in Modeling β€” The Consortium for Mathematics and its Applications, May 2024 (Team Award)
  • Second Prize, National College Student Computer System Ability Competition β€” National Association of Computer Education in Colleges and Universities, Aug 2025 (Team Award)
  • First Prize, Physics Competition β€” Beijing Physical Society, Dec 2023
  • AVIC Scholarship β€” Aviation Industry Corporation of China
  • BUAA Academic Excellence Scholarship, First Prize β€” Beihang University
  • Competition Scholarship, Top Prize β€” Beihang University
  • Innovation and Entrepreneurship Scholarship, Second Prize β€” Beihang University
  • Outstanding Student Leader of Beihang University (2023, 2024, 2025)
  • Outstanding League Cadre of Beihang University
  • Outstanding Student of Beihang University

πŸŽ“ Education

  • 2026 - now, Ph.D. Student, Electronic Design Automation and Computing Systems, School of Integrated Circuits, Peking University
  • 2022 - 2026, B.Eng. in Computer Science and Technology, School of Computer Science and Engineering, Beihang University

🀝 Service

  • Tutorial Chair, APPT
    T2: Graph.hls: A Compiler Framework for Composable Graph Accelerator Design

πŸ‘¨β€πŸ« Teaching

  • Advanced Teaching Assistant, Computer Organization, Beihang University, 2025/1-2025/12
  • Teaching Assistant, Operating Systems, Beihang University, 2024/09-2025/06
  • Teaching Assistant, Computer Organization, Beihang University, 2024/07-2024/12

🧰 Skills

  • Programming: Python, C++, CUDA, SystemVerilog
  • Systems: LLM serving, distributed inference, processor design, reinforcement-learning experiment platforms
  • English: CET-6